Design: Weekly Summary (January 19-25, 2026)

I kept a loose eye on design talk this week. It wasn’t one loud chorus. It was more like a cluster of small conversations in different rooms. Some were grumpy. Some were wistful. Some were trying hard to guess the future. I would describe them as a mix of hands-on nitpicking and bigger-picture worrying. To me, it feels like people are wrestling with the same thing: how do we keep making things that actually help, and not just add noise?

Quick sketch of where the noise came from

There were a few clear threads. AI and tools vs human craft. Icons and the tiny things that betray bigger shifts. Process — whether the old rituals still help or only slow you down. And nostalgia, not as a cute habit but as a shorthand for values people miss. You’ll see each of these pop up in the posts I read.

I’d say the week felt a bit like cleaning out a kitchen drawer. You find a helpful little gadget. You find a broken spoon. You talk for five minutes about whether you should throw the whole drawer out, or tidy it properly. There are practical decisions under debate. Small choices with outsized consequences.

AI trying to do the job — but not quietly taking over

There was a fair bit of skepticism about AI’s current limits. George Saines wrote about trying to design a logo with AI. He’s a former product designer, so he did more than poke a few buttons. He tried prompts. He tried different tools. He tuned things. He didn’t get a logo he felt could ship. That stuck with me.

I would describe his findings as: AI can throw up ideas. It can make a collage of plausible shapes. But it doesn’t yet make the kind of decisions a person would, when the logo has to work at 8px, on a bad screen, in two colours, in print, on a bus billboard, and so on. The post doesn’t slam the door on AI. It just says the door is still ajar. You’d need a human to walk through.

That feeling echoes in another piece. Maggie Appleton took on the chaos of agent orchestration in software with a critique of the Gas Town project. Her tone is part-analyst, part-finger-on-the-map. She warns about the cost of letting a bunch of agents run without a clear plan. The technical details are interesting if you care about infrastructure and dev teams, but the design lesson is what I kept thinking about: automation without guardrails is just delegation without responsibility.

To me, it feels like people are building factories where the machines talk to each other and hope the product leaves neatly packaged. That almost never happens. You need layout and a foreman.

Both George Saines and Maggie Appleton are, in different ways, reminding designers that tools don’t replace judgment. They change the work, sure. But they also create new failure modes.

Process: trust the prototype, not the ritual

There’s a short, sharp post from Simon Willison that reads like a warning: don’t cling to the old design process as some talisman. He says, and I’d paraphrase, that prototyping matters more now. Quick loops that show what actually works are better than long documents that try to predict every corner case.

I’d say that’s a useful nudge. It’s easy to pray at the altar of process. I’ve seen teams do that. It feels safe. But Willison points out the obvious but overlooked: when your tools can try things fast, your process should too. The danger is human inertia. People keep following rituals they used before the world got faster.

There is a nice cross-link with the AI discussion. If AI can spit out many variants quickly, prototyping becomes experimental in a different way. But you still need human choices. You still need a person to say which variant survives. The human role is curating, editing, and sometimes saying no.

Tiny things that tell you everything — icons and UI micro-decisions

Two posts had a duel about macOS Tahoe icons and small UI choices. Cal Henderson went off on how the icons feel worse. He used a strong word about the general decline in quality — a kind of 'enshittification', he calls it — and that whole idea resonated. Small things adding up to a big less-human feel.

Then Lucio Bragagnolo pushed back. He said the problem is being exaggerated. In his view, the column view issue is solvable by resizing and it only bites in certain setups. So, arguments about decline vs adaptability showed up in miniature.

To me, this duel is important because icons are where design makes contact with daily life. An icon is like a mug on your desk. If it’s chipped, you notice. If it’s beautiful, you maybe don’t notice consciously but you feel different reaching for it every morning. That’s the odd power of interface micro-decisions.

A small detour: I’m reminded of how you notice the wrong smell in a kitchen. It’s not the whole house collapsing. It’s just one thing out of place and your brain flags the whole space as off.

Continuity, lineage, and stealing from the best

There’s a very human story in a short note from daveverse about icons and historical continuity. He took the mailbox flag idea from WordLand II to reveal OG metadata, and pointed back to Susan Kare’s little Macintosh desk accessories like the Alarm Clock.

He doesn’t just admire pixels. He’s saying: design is a conversation across time. You can nick the good bits. Steal from the best, he says. I’d say that is less theft, more inheritance. Design often improves when people riff on older ideas. It’s like music sampling. You do it with respect or you botch it.

I liked that because it feels reverent without being stuck in the past. He’s reminding us that icons are shorthand. They carry a memory. That’s why a good icon feels obvious and wrong to mess with lightly.

Architecture and emotional design — the big quiet pieces

Not all posts were about screens and AI. Christopher Schwarz wrote about Frank Lloyd Wright and E. Fay Jones, and their ability to make buildings that felt like they belonged to the place. He told a story about growing up and being moved by Thorncrown Chapel. The piece reads as personal memory and a lesson about design that is patient and site-aware.

This one slid in quietly among the tech noise and it matters. Architecture reminds us that good design isn’t always flashy. Sometimes it’s the way a building lets you breathe. I would describe Wright and Jones’s work as a kind of slow UX. They thought about light, materials, and the human scale. That’s design thinking with a lot of patience.

It’s an important counterpoint to the rush of UI and AI. Not everything needs to be iterated in a week. Some things need to be lived in.

Publishing, social media, and the decline of meaning

There was a link roundup from Nicolas Magand that stitched together thoughts about algorithms, publishing, and a new blog design by Robin Rendle. It wasn’t a manifesto. It was a hand of links and a few sharp notes about how social media’s attention economy is reshaping what counts as meaningful publishing.

The phrase that hung with me was that social platforms make publishing feel like shouting in a crowded room. The room is noisy. The mic is hot for a bit. The result: people chase spikes instead of slow conversations. The result reaches even into design. You start optimising for a quick tap rather than a durable tool.

I’d say link roundups have a tiny, forgotten utility. They act like a local paper’s noticeboard. You can get one meaningful lead from a cluster of links. Magand’s list nudged me to older ideas and new ones. Worth a click if you like following tangents.

A tiny fight about what’s broken and what’s fine

Back to macOS for a second. The disagreement between Cal Henderson and Lucio Bragagnolo is kind of a microcosm of a larger pattern. One person sees a design trend as symptomatic of decline. The other sees the same thing and thinks it’s manageable or exaggerated.

I think both are right in part. That’s human. Sometimes change is sloppy and needs pushing back. Other times the squeaky wheel gets oil and things are fine. The important bit is that people are arguing. That’s actually healthy. If everyone shrugged, it would be worse. Here, the argument is practical. It’s not just aesthetic sniping.

Design bottlenecks and the human cost of automation

Maggie’s piece circles back to another worry. When you push more intelligent tools into a workflow, you create bottlenecks in other places. Her critique of Gas Town’s architecture showed how easy it is to get tangled in complexity that looks clever but costs time and money.

I would describe the risk as: you trade one kind of friction for another. The fancy agent system removes some repetitive work. But now you have to design, debug, and pay for coordination. Teams can lose sight of who owns what. That’s not theory. That’s payroll and late-night bug hunts.

There’s also a social thing here. Automation sounds attractive because we imagine less work. But often people end up firefighting. The work doesn’t disappear. It moves.

A short detour about craft vs output

This week’s posts reminded me that design sits between craft and production. George Saines wanted a logo that felt right. He wanted something that would live and breathe with a brand. AI pushed back with lots of variants. It’s like ordering a tailcoat from a vending machine. You get something that looks like a tailcoat, but the fit is wrong.

Meanwhile, in systems and architecture, people are tempted to replace tailoring with mending machines. That’s fine until you notice the seams. You still need people who know how to hem.

The small comforts and the human stories

A few posts leaned into personal story. The architecture piece that mentioned a family event at Thorncrown Chapel. The note about Susan Kare and the old Macintosh desk accessories. These are small human anchors.

I’d say these are crucial. When design discussion gets abstract — 'interfaces', 'agents', 'publishing mechanics' — these personal notes re-ground it. They remind you why anyone cares. Design touches the things people live with. It hits home.

There’s nothing hokey about that. Design is not a spreadsheet metric. It’s the way light hits a pew. It’s the way an icon makes you pause or move on. If you lose that human angle, you can still build features. But you lose the reason people keep using what you made.

Little patterns that popped up

• A healthy distrust of AI as a finished craftsman. People expected more polish from the tools, and were disappointed. That disappointment feels specific. It’s about practical outputs like logos and reliable agent systems.
• A debate about micro-changes vs macro-signals. Are ugly icons a sign of decline, or just noise? Different readers answer differently. Both sides had reasonable points.
• A push toward prototyping and quick feedback loops. The old design process is getting questioned where machines let you try things fast.
• Nostalgia as evidence, not just sentiment. References to Susan Kare, Wright, and Jones were used to argue for continuity and quiet strength. It’s not about romanticising the past. It’s about learning small, repeatable good choices.

How the week made me think about what to read next

If you like hands-on critique, read George Saines on logos. If you want a systems-level eye on where agentic work can go sideways, read Maggie Appleton. If you want the small, exact gripe about icons, read Cal Henderson and then the reply from Lucio Bragagnolo to see the other side.

If you want a slightly calmer, slower piece, Christopher Schwarz gives architecture as an example of patient design. daveverse is a quick, bright note about continuity in icons. And Simon Willison is the short nudge about changing processes.

There’s a link pile from Nicolas Magand if you want to wander further. I’d say it’s like a friend leaving a few books on your kitchen table. You might pick through them.

Some slightly messy thoughts

Design people are split between engineers who want to optimise and craftsmen who want to protect standards. The week’s posts read like that split in dialogue. There’s no tidy resolution. There rarely is. But the friction is useful. It forces questions about what we value.

I kept finding myself going back to one metaphor: design is like cooking for guests you haven’t met. Sometimes you have a recipe. Sometimes you improvise with what’s in the fridge. Tools can chop faster, but they don’t taste the soup and tell you if it needs salt. There’s always a last bit of human sense-making.

Another small thought: we talk about automation as if it’s a single event. It’s not. It’s a long tail of decisions. Agents and AI create new layers to manage. That’s the real design challenge now. It’s less about magic and more about orchestration.

A few examples that stuck

• The logo experiments. They were a kind of laboratory. Lots of outputs. Little that felt ready. The lesson: AI gives you sketches, not signed prints.
• Gas Town. An interesting experiment. Messy in places. A reminder that coordination costs money and attention.
• Icons in Tahoe. A tiff that shows why small things matter. You can fix columns or you can fix attention to small details. Which one do you pick?
• Thorncrown Chapel. A reminder that light and placement beat flashiness on occasion.

If you want to dive deeper

Click into the posts. I’d recommend starting with the ones that most closely match your current frustration. Want practical design limits of AI? Start with George Saines. Want to think about agents and scale? Maggie Appleton is a good doorway. Into the macOS grumble? Read Cal Henderson and then the counterpoint by Lucio Bragagnolo.

Also, if you like to chase threads, Nicolas Magand left a string of links that send you sideways. They’re those rabbit holes that end in reading something clever in the middle of the night.

I’ll leave it there, because the posts themselves have the color and the examples. This week the discussion felt a little like a neighbourhood meeting. People shouted, people worried, people brought pie. Some arguments were old. Some were new. The parts that mattered were the specific fixes and the concrete complaints. That’s the kind of design talk worth following. It makes you want to open the drawer and sort things out.